Bayesian bias modelling for causal inference in statistics

点击翻译按钮获取中文摘要

Causal inference is a thriving area of innovation in statistics, with important applications in the social sciences, medicine, and other fields. The goal of this research program is to develop new statistical methodology for Bayesian bias analysis in causal inference. The research will build on the recent progress of the applicant, which includes new causal inference methods that have been published in statistics journals, and additionally, that have had an impact in domain-specific applied areas. The short-term objectives are A) To develop a Bayesian method to explore sensitivity to multiple unmeasured confounders in causal inference, and B) to develop and investigate novel Bayesian methods for estimating propensity scores to control confounding. Longer-term goals include C) the development of Bayesian approaches to model heterogeneity of bias in the context of meta-analysis of observational studies, and D) To study the role of Bayesian techniques to model bias from non-ignorable missing data in longitudinal studies. A key component of the research will be to study the performance of new Bayesian methods compared to standard frequentist approaches. The methodological approach will include computer simulation experiments, analytical results based on simple models, and a study of performance when applied real datasets. Trainees will be involved in all aspects of the research and gain knowledge in biostatistical methods applied to unique datasets. The proposed work will impact research on new methods for causal inference. Furthermore, it will impact how practitioners analyze data. Crucially, my academic appointment is located in the Faculty of Health Sciences at SFU, and I am an associate member in the Department of Statistics and Actuarial Sciences. This unique setting is a source of interdisciplinary training for students, new datasets, and inspiration for new statistical methods.